Let’s take a look at the total revenue per apu for the original revenue, the revenue after increasing the basic education component by 4%, and the revenue after increasing the entire formula by 4%.
The maps below provide the total revenue per APU for each scenario and the shades of color are “binned” by quintile - meaning each bin has 20% of the total public school districts, which is between 65 and 66 districts per bin.
The first thing that jumps out from the maps below is that our most rural districts, particularly along the borders of the state, receive the highest revenue per APU, no matter which scenario.
What we are looking for if there are any shifts/patterns in districts that receive the highest revenue per APU over the course of five years for each scenario. Glancing at these maps shows that there is very little change.
Next, let’s see if there are any patterns or relationships in these increases depending on type of school, it’s RUCA category, and regions. The charts below provide each school district’s total revenue per APU for each scenario, a boxplot for each scenario, as well as the relationship between type of school, RUCA category, and region to it’s total revenue per APU.
##
## Welch Two Sample t-test
##
## data: Original by Definition
## t = 6.1696, df = 289.41, p-value = 2.302e-09
## alternative hypothesis: true difference in means between group Charter Schools and group School Districts is not equal to 0
## 95 percent confidence interval:
## 510.6802 989.1443
## sample estimates:
## mean in group Charter Schools mean in group School Districts
## 8938.315 8188.403
##
## Welch Two Sample t-test
##
## data: Basic education only by Definition
## t = 6.0588, df = 283.55, p-value = 4.353e-09
## alternative hypothesis: true difference in means between group Charter Schools and group School Districts is not equal to 0
## 95 percent confidence interval:
## 563.8252 1106.4691
## sample estimates:
## mean in group Charter Schools mean in group School Districts
## 10309.482 9474.335
##
## Welch Two Sample t-test
##
## data: Entire formula by Definition
## t = 5.5696, df = 287.17, p-value = 5.87e-08
## alternative hypothesis: true difference in means between group Charter Schools and group School Districts is not equal to 0
## 95 percent confidence interval:
## 508.2686 1063.8400
## sample estimates:
## mean in group Charter Schools mean in group School Districts
## 10443.330 9657.276
## Df Sum Sq Mean Sq F value Pr(>F)
## Dem_Desc 3 33217295 11072432 7.277 8.73e-05 ***
## Residuals 504 766887828 1521603
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## Dem_Desc 3 42178336 14059445 7.308 8.36e-05 ***
## Residuals 504 969560524 1923731
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## Dem_Desc 3 4.296e+07 14319062 7.119 0.000109 ***
## Residuals 504 1.014e+09 2011492
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## planning.region 5 51572812 10314562 6.917 2.93e-06 ***
## Residuals 502 748532312 1491100
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## planning.region 5 66249362 13249872 7.035 2.28e-06 ***
## Residuals 502 945489499 1883445
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## planning.region 5 63091368 12618274 6.375 9.39e-06 ***
## Residuals 502 993657681 1979398
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## edr 12 80441079 6703423 4.611 4.16e-07 ***
## Residuals 495 719664044 1453867
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## edr 12 102875057 8572921 4.669 3.21e-07 ***
## Residuals 495 908863804 1836088
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Df Sum Sq Mean Sq F value Pr(>F)
## edr 12 101489568 8457464 4.383 1.15e-06 ***
## Residuals 495 955259481 1929817
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Let’s see if any of the districts shift in the quantiles going from the original to each of the scenarios.
The maps below highlight the districts that shifted quintiles from their original revenue quintile “rank”. There are very few that change.
When increasing the basic education component only, 8 schools shift - 4 move up one quintile while 4 others move down.
When increasing the entire formula 14 districts shift - 7 move up and 7 move down one quintile rank.
Next, let’s just see if the ranks change much across the state. The maps below provide the public school district ranks for each scenario. The darker the shade, the higher the rank meaning the higher the total revenue per APU. The maps show that the highest ranks are in northern Minnesota and southwest with significantly less around central Minnesota and the suburbs. I was hoping to see a pattern or change in ranks across the scenarios but that doesn’t seem to be the case.
## [1] "District Name" "number"
## [3] "type" "total.rev.original"
## [5] "total.rev.bsrevonly" "total.rev.entire.formula"
## [7] "group" "County Name"
## [9] "countyfp" "Dem_Desc"
## [11] "planning.region" "edr"
## [13] "awadm22" "total.rev.original.apu"
## [15] "total.rev.bsrevonly.apu" "total.rev.entire.formula.apu"
## [17] "scenario" "rank"
## [19] "geometry"
Perhaps looking at the maps through the lens of change. The maps below show the change in rank among all schools when going from the original total revenue per APU to the new values under each scenario. Shades of blue means a districts rank improved while yellow and orange means their ranks worsened.
When comparing the two maps, it shows that districts in southwest Minnesota improve in rank when the entire formula increases. When only the basic education revenue is increased, the central lakes region improves significantly.
However, this is all pretty subtle.
Even though there doesn’t seem to be much change or shift when comparing revenue per APU across all districts, the percentage change might highlight some patterns.
The maps below show the percent increase from the FY22 total revenues for each scenario and they show a pretty clear pattern.
When increasing the basic education component only, the highest percent increases occur in districts with higher property values such as the central lakes regions and the suburbs. However, if the entire formula increases, the highest percentages are in districts with lower property values such as central, western, and southern Minnesota.
Let’s see if there are any relationships in the amount of percent change from the original by the type of school, RUCA category, and regions. The charts below provide the pct change in total revenue for each school district for each scenario as well as the boxplots for these changes, and quick t-test and anova tables to test for significant differences.
Charter vs. Public: Both charts show that charter schools are more bunched together for each scenario and have very similar pct changes whereas the public schools have significantly more variance. In addition, there’s a huge difference in the gap between public vs charter when increasing the entire formula. The median for public schools in that scenario is 18% vs. 16.8% for charter schools.
RUCA: The RUCA chart shows considerable more low outliers among town/rural mix schools and urban/town/rural mix schools. The box plots show an interesting pattern. When the basic education component is the only thing increased, the median percentages and the middle 50% of values are relatively the same across all RUCA categories. However, if we increase the entire formula, entirely urban is considerably lower than the other categories. The median for the entirely urban is 17% vs. 18% for the other schools. In addition, the ANOVA test shows a significantly larger difference when increasing the entire formula vs. the basic education component only.
Planning Region: The planning region chart shows a similar pattern to the RUCA chart. When increasing the entire formula, the seven county metro is quite a bit lower than schools in the other planning regions. The median for the seven county metro is 16.9% vs. 17.5% to 18.2% for the other regions. In fact, the ANOVA test shows that there is no statistically significant difference when increasing the basic revenue component only vs. increasing the entire formula where there is a signficant difference.
EDR: Again, there is a significant decrease in the seven county metro compared to the other EDRs. In addition, there is much lower p-value for the differences across increasing the entire formula.
##
## Welch Two Sample t-test
##
## data: Basic education only by Definition
## t = -7.9365, df = 494.98, p-value = 1.402e-14
## alternative hypothesis: true difference in means between group Charter Schools and group School Districts is not equal to 0
## 95 percent confidence interval:
## -0.004955169 -0.002988592
## sample estimates:
## mean in group Charter Schools mean in group School Districts
## 0.1536963 0.1576682
##
## Welch Two Sample t-test
##
## data: Entire formula by Definition
## t = -28.774, df = 421.33, p-value < 2.2e-16
## alternative hypothesis: true difference in means between group Charter Schools and group School Districts is not equal to 0
## 95 percent confidence interval:
## -0.01201124 -0.01047517
## sample estimates:
## mean in group Charter Schools mean in group School Districts
## 0.1684829 0.1797261
## Df Sum Sq Mean Sq F value Pr(>F)
## Dem_Desc 3 0.000455 1.515e-04 3.754 0.011 *
## Residuals 504 0.020344 4.037e-05
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 6 observations deleted due to missingness
## Df Sum Sq Mean Sq F value Pr(>F)
## Dem_Desc 3 0.00540 1.80e-03 37.59 <2e-16 ***
## Residuals 504 0.02413 4.79e-05
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 6 observations deleted due to missingness
## Df Sum Sq Mean Sq F value Pr(>F)
## planning.region 5 0.000311 6.229e-05 1.526 0.18
## Residuals 502 0.020487 4.081e-05
## 6 observations deleted due to missingness
## Df Sum Sq Mean Sq F value Pr(>F)
## planning.region 5 0.008272 0.0016544 39.07 <2e-16 ***
## Residuals 502 0.021259 0.0000423
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 6 observations deleted due to missingness
## Df Sum Sq Mean Sq F value Pr(>F)
## edr 12 0.000952 7.937e-05 1.98 0.0242 *
## Residuals 495 0.019846 4.009e-05
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 6 observations deleted due to missingness
## Df Sum Sq Mean Sq F value Pr(>F)
## edr 12 0.008887 0.0007406 17.76 <2e-16 ***
## Residuals 495 0.020644 0.0000417
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 6 observations deleted due to missingness
To get a clear picture, lets see what districts shift up and down in their quintile rank when comparing the scenario that increases the entire formula vs. only increasing the basic education component.
The chart below shows that there is quite a bit of movement in terms of the quintile rank in pct change between the two scenarios. When comparing quintile ranks from the basic education only scenario with increasing the entire formula, districts in the lakes regions and suburbs drop between 2 and 4 ranks in their quintile while property poor schools increase a similar amount of ranks.
So now we need to figure out which components drive the change in total revenue. The chart below plots each school along with the percentage each component composes of the difference between increasing the basic education component only vs. increasing the entire formula. The chart shows that the components that are the primary drivers are;
The remaining categories do also have influence but across less schools.
Next, let’s see if there are any relationships between the type of school, it’s RUCA category, and it’s region. Since there are only a few components that make the biggest difference, we are going to include the following components;
The charts below provide the percent of each component of the formula comprises the difference between the total increase from increasing only the basic education component to increasing the entire formula for each district. Box plots are also included to see the difference in median values and concentrations.
Charter vs. public: the important piece here is that charter schools are not eligible to receive local optional aid, extended time revenue, or sparsity revenue. The biggest difference is the local optional aid received by public school districts. This category had a median value of 55% - meaning 55% of the difference between the two scenarios was made up by local optional aid. Charter schools typically had a higher percentage of operating capital aid, small schools revenue as well as English learners.
RUCA: Schools outside of entirely urban counties had significantly higher percentages in local optional aid, while more urban districts had higher percentages in operating capital aid, small schools revenue (likely charter schools), and English learners.
Planning Region: Interestingly, local optional aid barely shows up in the seven county metro. And, Northeast has a lot of districts at or near 0 as well. Other regions have the highest percentages of all components in the category. Operating capital is highest in the seven county metro. Small schools revenue is large middle 50%.
Let’s see what this looks like on a map (which be only public schools). The maps below show the top 4 categories that drive the difference between the two scenarios. It clearly shows that nearly all of greater Minnesota benefits from the local optional aid. This component drives between 40 and 75% of the difference between the two scenarios. For the seven county metro, operating capital becomes the highest. In some districts in northern Minnesota, small schools revenue and sparsity revenue are the largest drivers.
The second largest component for much of greater Minnesota districts is operating capital aid, small schools revenue, and English learners.